Nonparametric estimation from length-biased data under competing risks

A new model for cross-sectional lifetime data is presented. The model is based on the length-bias assumption, and it is adapted to situations in which several types of censoring may occur. The NPMLE of the survival function is derived. An EM-algorithm to approximate the NPMLE is devised. The performance of the introduced estimator is investigated through simulations. A real set of data collected as part of a study on unemployment duration in Spain is used for illustration purposes.